AI4DeSci is an ecosystem that is dedicated to continuously building sustainable infrastructure to support AI-empowered open & collaborative decentralized science. The ecosystem mission aims to build infrastructure that accelerate DeSci, to develop use cases of scientific impact, and promote DeSci through actions. The ecosystem aims to reinforce and facilitate both the Ideation and Actions that happen throughout DeSci project lifetime. This will manifest through the Collaborative Cloud and Science Crowdsourcing main platforms that are built together with community. The collaborative cloud allows scientist, researcher, students, and general public to leverage the decentralized independent cloud infrastructure offered for hosting collaboration. Collaborations that happen can be as simple as writing research paper together, as well as developing analysis protocol, or data processing pipeline. The Science Crowdfunding provide utilities for posting and organizing scientific task distribution campaign.
**Vision: ** Sustainable Infrastructure for AI-Empowered Open & Collaborative Decentralized Science
Mission:
[1] Build Infrastructure for Accelerating DeSci
[2] Develop AI4DeSci Use-Cases with Scientific Impact
[3] Promote DeSci Through Innovative Model and Actions
Build Exclusively on Akash.network
Akash.network
Problems we are solving
Cloud collaboration and Science Crowdsourcing are critical for Decentralized Science. Unfortunately, today the infrastructure for these activities are still marginal and mostly controlled by centralized big-tech entity, high-cost paywalled system, and often provided exclusively only for specific groups of people.
We believe decentralized science should and can benefit more with open-sourced infrastructure developed natively on Decentralized Supercloud like Akash.
Target user Groups Scientist, Engineers, Researchers, Teacher, Graduate Students, Science Enthusiast, ResearchPreneurs, and more…
Demo Video
For those who are wondering the video is recorded using an amazing app called mmhmm. Link to original video here
*Pitch Deck and Litepaper are available as .pdf on GitHub repository. *
Our Proposed Solution
Throughout this one month of Akashathon, we have built the following Proof-of-Concept and MVP
Check on the Official Website for AI4DeSci and start exploring the ecosystem proof-of-concept we built. Proudly 100% deployed all across Akash.network
*Expand below to see the rest of the Demo screenshot. Test it live through the corresponding public links or raw .ingress link.
**Click and expand below to see the public credentials for Akashathon **
Observed Challenges with Production Deployments on Akash Ecosystem
Throughout the Akashathon journey we also uncover severals pain points that will need to be alleviated later to support better deployment and more use cases on top of Akash.network.
- Monitoring the health and status of distributed deployments is a challenge without integrated dashboard for status reporting.
- Deployment instability when using reverse proxy is still questionable owing to the discrepancy in network performance of different provider, e.g. static, and dynamics, IP, TLS certificates, etc.
- Persistent storage options reading often faulty and not available on most provider, which is seems due to error in attribute matching and not due to resource limitation.
- more details will be properly documented and raised as issue or feature request through github …
What's next for AI4DeSci
Some more details for those interested to learn more ...
Upcoming Core Features Toward Decentralized HPC on Akash : Start from proper slurm –jupyter integration for hub.ai4desci.com to have worker node across Akash Deployments
Modular mini-app builder and Automated Packaging + Deployment: Utilizing streamlit components and label-studio SDK to allow for self-customization and auto-deployment of AI4DeSci mini-app on app.ai4desci.com
*Real-World Use Case Quick Prototyping: * Work closely with ResearchHub team and community in developing Real-world DeSci use collaboration use-case.